Higher order motion blur may be required to faithfully capture object rotations and long-exposure renderings. Higher order motion blur comes at a significant cost. The sample coverage test is more expensive, and a large number of samples per pixels (distributed in time) are needed to carefully capture non-linear motion effects. It is, therefore, desirable to reduce the number of sample tests. Unfortunately, previous algorithms for linear per-vertex motion blur have low culling ratios for highly non-linear motion.